Summary
Unstandardized parameter estimates (what we call "map optcom betas" in generate_metrics()) are scaled by the variance of the data, which means that the parameter estimate maps will be confounded by the scale of the data. I don't think it makes sense to calculate dependence metrics, such as "variance explained" (more appropriately called coefficient energy), "dice_FT2", and "dice_FS0", with maps that have that confound in them. Instead, I propose that we use the standardized parameter estimate maps added in #1332 for these measures.
I do think we probably want to use the unstandardized maps for figures, but I could be wrong about that. (EDIT: I tried showing the standardized maps in the figures and they looked terrible, so I'm happy with using the unstandardized ones)
Additional Detail
Related to #1309, #1315, and #1332.
Summary
Unstandardized parameter estimates (what we call "map optcom betas" in
generate_metrics()) are scaled by the variance of the data, which means that the parameter estimate maps will be confounded by the scale of the data. I don't think it makes sense to calculate dependence metrics, such as "variance explained" (more appropriately called coefficient energy), "dice_FT2", and "dice_FS0", with maps that have that confound in them. Instead, I propose that we use the standardized parameter estimate maps added in #1332 for these measures.I do think we probably want to use the unstandardized maps for figures, but I could be wrong about that. (EDIT: I tried showing the standardized maps in the figures and they looked terrible, so I'm happy with using the unstandardized ones)
Additional Detail
Related to #1309, #1315, and #1332.